The TRANSFORM Patient Safety Project: A Microsystem Approach to Improving Outcomes on Inpatient Units Clarence H. Braddock III MD, MPH1, Nancy Szaflarski, PhD, RN2, Lynn Forsey, PhD, RN3, Lynn Abel, MSN, RN4, Tina Hernandez-Boussard, PhD, MPH5, and John Morton, MD, MPH5 1
Department of Medicine, David Geffen School of Medicine University of California, Los Angeles, CA, USA; 2Quality & Effectiveness Department, Stanford Health Care, Stanford, CA, USA; 3Department of Nursing, Mills-Peninsula Health Services, Burlingame, CA, USA; 4Patient Care Services, Stanford Health Care, Stanford, CA, USA; 5Department of Surgery, Stanford University School of Medicine, Stanford, CA, USA.
BACKGROUND: Improvements in hospital patient safety have been made, but innovative approaches are needed to accelerate progress. Evidence is emerging that microsystem approaches to quality and safety improvement in hospital care are effective. OBJECTIVE: We aimed to evaluate the effects of a multifaceted, microsystem-level patient safety program on clinical outcomes and safety culture on inpatient units. DESIGN: A 1-year prospective interventional study was conducted, followed by a 6-month sustainability phase. SETTING AND PARTICIPANTS: Four medical and surgical inpatient units within an academic university medical center were included, with registered nurses and residents representing study participants. INTERVENTIONS: In situ simulation training; debriefing of medical emergencies; monthly patient safety team meetings; patient safety champion role; interdisciplinary patient safety conferences; recognition program for exemplary teamwork. OUTCOMES: Hospital-acquired severe sepsis/septic shock and acute respiratory failure; unplanned transfers to higher level of care (HLOC); weighted risk-adjusted mortality. Safety culture was measured using a widely accepted, validated survey. RESULTS: Rates of hospital-acquired severe sepsis/ septic shock and acute respiratory failure decreased on study units, from 1.78 to 0.64 (p=0.04) and 2.44 to 0.43 per 1,000 unit discharges (p=0.03), respectively. The mean number of days between cases of severe sepsis/septic shock increased from baseline to the intervention period (p=0.03). Unplanned transfers to HLOC increased from 715 to 764 per 1,000 unit transfers (p = 0.08). The weighted risk-adjusted observed-to-expected mortality ratio on all study units decreased from 0.50 to 0.40 (p 2.0 mmol/L • absent bowel sounds • platelet count < 100,000 • total serum bilirubin > 4 mg/dl • mottled or cool extremities • systolic blood pressure < 90 mmHg or decrease of 40 mmHg or more from patient’s hospital baseline B. Septic Shock (ICD-9-CM: 785.52) Clinical Definition: Severe sepsis with hypotension (as defined above) that did not resolve with two liters of intravenous fluid boluses or which reoccurred after administration of two liters. C. Acute Respiratory Failure [ICD-9-CM: 518.81 with either 96.04 (intubation) or 93.90 (noninvasive ventilation)] Clinical Definition: One or more of the following: • arterial oxygen saturation of less than 85 % • arterial partial pressure of oxygen of less than 60 mmHg • arterial partial pressure of carbon dioxide of greater than 65 mmHg and, which resulted in either emergent use of bi-level positive airway pressure or intubation on a study unit
1. In Situ, High-Fidelity Simulation Training. Our goal was 90 % participation in at least one simulation exercise during the intervention. We conducted four in situ simulation exercises per study unit per month, on both day and night shifts. Simulation training and debriefings were facilitated by UBMDs and CNSs using a standardized checklist. 2. Debriefing of Medical Emergencies. Charge nurses on study units were responsible for conducting debriefings following RRT and Code Blue calls to discover factors contributing to the call. Our goal was to have 90 % of calls debriefed. 3. Unit Patient Safety Champion Role. At least one registered nurse per shift on each unit functioned as a patient safety champion. 4. Monthly Unit Patient Safety Team Meetings. Unit leaders were responsible to hold a meeting to review a case involving interdisciplinary care issues, as well as to define an action plan and monitor improvement. Our goal was 90 % leader participation in monthly meetings. 5. Quarterly Interdisciplinary Patient Safety Conference. A conference was to be held involving nurses, residents and attending physicians to discuss and improve upon interdisciplinary teamwork or care issues. Our goal was four conferences per year. 6. Individual Performance Recognition. An award was to be given to a nurse or resident nominated for demonstrating exemplary teamwork. Our goal was one award per month.
Sustainability Period Based on the success of Project TRANSFORM after 1 year of intervention, we sustained the project for an additional 6 months. All study interventions were maintained, though simulation training decreased to one exercise per month.
Outcome Measures and Data Collection Administrative (ICD-9-CM) codes were used to identify patients who had a study complication not present on hospital admission (Table 2). A chart abstractor retrospectively reviewed records of identified cases to confirm that: (1) the complication developed while the patient was on a study unit, (2) the hospital length of stay (LOS) on the study unit was greater than 12 h, and (3) complications met the clinical case definition (Table 2). Twenty-five cases of each complication were randomly selected to determine inter-observer agreement. Medical records of selected cases were audited by an author (NS) using
the clinical definition for each complication. We found 96 % agreement among cases of severe sepsis/septic shock (kappa coefficient 0.9495, Z 8.50) and 97.3 % for acute respiratory failure (kappa coefficient 0.9657, Z 8.55). A hospitalist with no involvement with the program validated all final identified complications. Rates of complications on study units were compared to all hospital, non-study inpatient units, which included medical and surgical wards and intermediate intensive care units (ICUs) whose patient populations were stable across the study periods. Complications were identified using administrative codes without medical record validation. Data of all patients transferred to a higher level of care (HLOC) were obtained from internal administrative data. The abstractor reviewed medical records to determine whether the transfer was unplanned and the medical condition necessitating the transfer. An unplanned transfer was defined as any: (1) transfer to an ICU or an intermediate ICU and was not scheduled for ICU or intermediate Intensive Care Unit (IICU) admission following elective surgery or an elective procedure; or (2) transfer to the operating room (OR) for emergent reasons (excluded elective and planned surgery). We excluded patients who had a “do not resuscitate” (DNR) order at the time of transfer. All hospital deaths that had at least one stay on a study unit during hospitalization were identified; deaths were excluded if: (1) a DNR order was documented during the first 24 h of admission and (2) the LOS on a study unit was less than 24 h. For a given patient, the total LOS spent on a study unit (s) was determined as a percentage of the total hospital LOS. The Elixhauser Comorbidity Index was used for risk adjustment.26,27 Both observed and expected mortality were then weighted in proportion to the total LOS on a study unit, and the “weighted” risk-adjusted observed-to-expected (O:E)
JGIM
Braddock et al.: The TRANSFORM Patient Safety Project
Table 3. Safety Issues Discovered through Program Implementation and Actions Taken Teamwork Issues Suboptimal communication of changes in patient status
Inadequate bedside identification of caregiver roles Staff nurses not reliably informing charge nurses of patient condition changes Delays in communicating patient condition changes due to mobile nature of staff nurses on unit Early Detection Issues Failure to compare current status to baseline status Delays in diagnosis and treatment by residents due lack of consultation with senior clinician Variability in knowledge of diagnosis and treatment of study complications
Role of “relief nurse” on unit was task-driven Care Escalation Issues Least experienced resident being called first for urgent patient situation Suboptimal knowledge of chain of command (COC) and assertion in escalating care
Actions • 5-min videos demonstrating optimal Situation-Background-Assessment-Recommendation (SBAR) communication by unit clinical leaders was shown to staff during daily huddles • Charge nurses and patient safety champions were trained to “ask for an SBAR” from nurses when hearing of a condition change • A nursing competency on SBAR created requiring 1:1 demonstration and rated by CNSs • SBAR competency was observed during simulation training; if suboptimal, required to repeat performance • Paging guideline was implemented on study units to enhance use of SBAR when communicating patient problems • Extender placed on name badges identifying role (e.g., nurse, resident, attending physician) • Role identification was observed during simulation training and 1:1 feedback given during debriefing • Charge nurses began rounding on patients minimally once a shift to prompt staff nurses to relay clinical changes • Competency of having nurses communicate changes to charge nurses was observed during simulation training and feedback given • Nurses taught to relay phone number in patient’s room when paging resident to ensure a more timely response • Nursing units purchased individual phones for nurses to avoid delays in communicating problems to physicians • Communication behavior observed during simulation training and 1:1 feedback given during debriefing Actions • Nurses taught how to reconfigure electronic medical record to view baseline vital signs and laboratory values for comparison • Education on need for early consultation was reinforced during simulation training, patient safety conferences and monthly meetings • Consultation competency for nurses and residents observed during simulation training and 1:1 feedback given • Sepsis order set was created in electronic medical record to support optimal ordering of diagnostic tests and treatment • Clinical guidelines for study complications were routinely discussed in depth at quarterly patient safety conferences • Knowledge competencies and ordering behaviors were observed during simulation training and feedback given • Role description of relief nurse was revised to emphasize coaching role to help nurses critically think through condition changes and take action Actions • Unit guideline defined new process that intern and senior resident need to be concurrently called for urgent/emergent situations • Guideline disseminated among nurses and supported by UBMDs in monthly unit resident orientation • CNSs reinforced COC to staff nurses during huddles • Patient safety conferences reviewed COC and method for asserting care escalation in face of perceived authority gradients • Patient safety champions empowered staff nurses to elicit COC • COC knowledge and assertion competencies observed during simulation training and feedback given
methodology.28 Participant responses of “agree” and “strongly agree” on the survey’s five-point Likert scale constituted the mean percent positive scores (0–100 % measurement range). All survey results are reported as percent positive scores.
mortality ratio was determined, based on the premise that clinical deterioration of patients is proportional across inpatient units where they stayed during the LOS. Clinical outcomes were measured at baseline (July 2009–June 2010) and for intervention and sustainability periods. Safety culture on study units was assessed at baseline and after 1 year, using the Agency for Healthcare Research and Quality’s (AHRQ) Hospital Survey on Patient Safety Culture (HSOPS).28 HSOPS data for nurses on non-study, medicalsurgical inpatient units were available for baseline and postintervention periods from an organizational-wide survey. Survey results were analyzed following AHRQ scoring
Statistical Analysis Multivariate regression was used to assess the effects of the program on complication rates and mortality over the three time periods, adjusting for patient characteristics. Control variables included patient age, race, payer status, inpatient
Table 4. Rates of Hospital-Acquired Severe Sepsis/Septic Shock and Acute Respiratory Failure on Study Units
Rate of Severe Sepsis/Septic Shock (complications/1000 unit discharges) Cases of Severe Sepsis/Septic Shock Rate of Acute Respiratory Failure (complications/1,000 unit discharges) Cases of Acute Respiratory Failure Unit Discharges
Baseline Period (July 2009–June 2010)
Intervention Period (July 2010–June 2011)
Sustain Period (July–Dec 2011)
Odds Ratio (95 % CI)
p value
1.78
1.21
0.64
0.53 (0.29–0.96)
0.04
16 2.44
11 2.10
3 0.43
0.58 (0.35–0.96)
0.03
22 9,000
19 9,058
2 4,685
JGIM
Braddock et al.: The TRANSFORM Patient Safety Project
admission status, and Elixhauser comorbidities.26,27 Model parameters were estimated using unconditional least squares and goodness-of-fit evaluated by the likelihood ratio test. G charts were used to calculate the number of days between complications. Linear regression was used to evaluate the effect of the program on the rate of unplanned transfer to HLOC. Pearson’s chi-squared tests with Yates’ continuity correction and Fisher’s Exact test were used to analyze the safety culture data. All statistical analyses were performed in SAS v9.3 (SAS Institute, Inc., Cary NC, USA). A p value of < 0.05 was considered to denote statistical significance for all outcomes.
Microsystem leaders led 90 % of monthly patient safety team meetings during intervention and sustainability periods, and each study unit maintained one unit patient safety champion per shift. Patient safety conferences were held each quarter and 90 % of RRT and cardiopulmonary arrest calls were debriefed. Each month during the study, one nurse or resident received teamwork recognition. Of the numerous safety issues discovered by microsystem leaders during the study, the most prevalent issues focused on interdisciplinary teamwork, early detection and care escalation (Table 3).
Outcomes RESULTS
Program Interventions After 1 year of intervention, 90 % of nurses (N=247) and 92 % of residents (N=56) on study units had participated in at least one simulation training exercise, with 43 % of nurses and 20 % of residents participating in two or more. Four simulation exercises were conducted each month on each unit during the intervention period and one exercise was conducted each month on each unit during the sustainability period. At the end of the sustainability period, 98 % of nurses (N=269) and 100 % of residents (n=61) had participated in simulation training.
A total of 13,743 patients were discharged from the study units during the study. The rate of hospital-acquired severe sepsis/ septic shock on study units decreased from 1.78 to 0.64 per 1,000 unit discharges across intervention and sustainability periods (odds ratio, 0.53; 95 % CI, 0.29–0.96; p=0.04) (Table 4). The mean number of days between cases of severe sepsis/septic shock statistically increased from 18 at baseline to 34 during the intervention period (p=0.03), with rates decreasing to 25 during the sustainability period (Fig. 1a). No significant change in the mean number of days between cases of acute respiratory failure occurred across study periods (p=0.25) (Fig. 1b). The rate of hospital-acquired acute respiratory failure decreased from 2.44 to 0.43 across the study periods
Figure 1. G-Charts showing days between cases on study units of: a severe sepsis/septic shock, and b acute respiratory failure.
JGIM
Braddock et al.: The TRANSFORM Patient Safety Project
(odds ratio, 0.58; 95 % CI, 0.35–0.96; p=0.03) (Table 4). While rates on study units decreased over time, the rates of both complications occurring on non-study inpatient units statistically significantly increased during intervention and sustainability periods (Fig. 2a and b). The rate of unplanned transfers to HLOC increased from 715 to 764 per 1,000 unit transfers across the study periods (p=0.08) (Table 5a). The medical conditions primarily contributing to the increase in unplanned transfers across the study periods were conditions not related to study complications (Table 5b). The weighted, risk-adjusted observed-to-expected (O:E) mortality ratio on study units was 0.50 at baseline, which decreased to 0.44 during the intervention period, and decreased to 0.40 during the sustainability period (odds ratio, 0.95; 95 % CI, 0.94–0.97; p
Department of Medicine, David Geffen School of Medicine University of California, Los Angeles, CA, USA; 2Quality & Effectiveness Department, Stanford Health Care, Stanford, CA, USA; 3Department of Nursing, Mills-Peninsula Health Services, Burlingame, CA, USA; 4Patient Care Services, Stanford Health Care, Stanford, CA, USA; 5Department of Surgery, Stanford University School of Medicine, Stanford, CA, USA.
BACKGROUND: Improvements in hospital patient safety have been made, but innovative approaches are needed to accelerate progress. Evidence is emerging that microsystem approaches to quality and safety improvement in hospital care are effective. OBJECTIVE: We aimed to evaluate the effects of a multifaceted, microsystem-level patient safety program on clinical outcomes and safety culture on inpatient units. DESIGN: A 1-year prospective interventional study was conducted, followed by a 6-month sustainability phase. SETTING AND PARTICIPANTS: Four medical and surgical inpatient units within an academic university medical center were included, with registered nurses and residents representing study participants. INTERVENTIONS: In situ simulation training; debriefing of medical emergencies; monthly patient safety team meetings; patient safety champion role; interdisciplinary patient safety conferences; recognition program for exemplary teamwork. OUTCOMES: Hospital-acquired severe sepsis/septic shock and acute respiratory failure; unplanned transfers to higher level of care (HLOC); weighted risk-adjusted mortality. Safety culture was measured using a widely accepted, validated survey. RESULTS: Rates of hospital-acquired severe sepsis/ septic shock and acute respiratory failure decreased on study units, from 1.78 to 0.64 (p=0.04) and 2.44 to 0.43 per 1,000 unit discharges (p=0.03), respectively. The mean number of days between cases of severe sepsis/septic shock increased from baseline to the intervention period (p=0.03). Unplanned transfers to HLOC increased from 715 to 764 per 1,000 unit transfers (p = 0.08). The weighted risk-adjusted observed-to-expected mortality ratio on all study units decreased from 0.50 to 0.40 (p 2.0 mmol/L • absent bowel sounds • platelet count < 100,000 • total serum bilirubin > 4 mg/dl • mottled or cool extremities • systolic blood pressure < 90 mmHg or decrease of 40 mmHg or more from patient’s hospital baseline B. Septic Shock (ICD-9-CM: 785.52) Clinical Definition: Severe sepsis with hypotension (as defined above) that did not resolve with two liters of intravenous fluid boluses or which reoccurred after administration of two liters. C. Acute Respiratory Failure [ICD-9-CM: 518.81 with either 96.04 (intubation) or 93.90 (noninvasive ventilation)] Clinical Definition: One or more of the following: • arterial oxygen saturation of less than 85 % • arterial partial pressure of oxygen of less than 60 mmHg • arterial partial pressure of carbon dioxide of greater than 65 mmHg and, which resulted in either emergent use of bi-level positive airway pressure or intubation on a study unit
1. In Situ, High-Fidelity Simulation Training. Our goal was 90 % participation in at least one simulation exercise during the intervention. We conducted four in situ simulation exercises per study unit per month, on both day and night shifts. Simulation training and debriefings were facilitated by UBMDs and CNSs using a standardized checklist. 2. Debriefing of Medical Emergencies. Charge nurses on study units were responsible for conducting debriefings following RRT and Code Blue calls to discover factors contributing to the call. Our goal was to have 90 % of calls debriefed. 3. Unit Patient Safety Champion Role. At least one registered nurse per shift on each unit functioned as a patient safety champion. 4. Monthly Unit Patient Safety Team Meetings. Unit leaders were responsible to hold a meeting to review a case involving interdisciplinary care issues, as well as to define an action plan and monitor improvement. Our goal was 90 % leader participation in monthly meetings. 5. Quarterly Interdisciplinary Patient Safety Conference. A conference was to be held involving nurses, residents and attending physicians to discuss and improve upon interdisciplinary teamwork or care issues. Our goal was four conferences per year. 6. Individual Performance Recognition. An award was to be given to a nurse or resident nominated for demonstrating exemplary teamwork. Our goal was one award per month.
Sustainability Period Based on the success of Project TRANSFORM after 1 year of intervention, we sustained the project for an additional 6 months. All study interventions were maintained, though simulation training decreased to one exercise per month.
Outcome Measures and Data Collection Administrative (ICD-9-CM) codes were used to identify patients who had a study complication not present on hospital admission (Table 2). A chart abstractor retrospectively reviewed records of identified cases to confirm that: (1) the complication developed while the patient was on a study unit, (2) the hospital length of stay (LOS) on the study unit was greater than 12 h, and (3) complications met the clinical case definition (Table 2). Twenty-five cases of each complication were randomly selected to determine inter-observer agreement. Medical records of selected cases were audited by an author (NS) using
the clinical definition for each complication. We found 96 % agreement among cases of severe sepsis/septic shock (kappa coefficient 0.9495, Z 8.50) and 97.3 % for acute respiratory failure (kappa coefficient 0.9657, Z 8.55). A hospitalist with no involvement with the program validated all final identified complications. Rates of complications on study units were compared to all hospital, non-study inpatient units, which included medical and surgical wards and intermediate intensive care units (ICUs) whose patient populations were stable across the study periods. Complications were identified using administrative codes without medical record validation. Data of all patients transferred to a higher level of care (HLOC) were obtained from internal administrative data. The abstractor reviewed medical records to determine whether the transfer was unplanned and the medical condition necessitating the transfer. An unplanned transfer was defined as any: (1) transfer to an ICU or an intermediate ICU and was not scheduled for ICU or intermediate Intensive Care Unit (IICU) admission following elective surgery or an elective procedure; or (2) transfer to the operating room (OR) for emergent reasons (excluded elective and planned surgery). We excluded patients who had a “do not resuscitate” (DNR) order at the time of transfer. All hospital deaths that had at least one stay on a study unit during hospitalization were identified; deaths were excluded if: (1) a DNR order was documented during the first 24 h of admission and (2) the LOS on a study unit was less than 24 h. For a given patient, the total LOS spent on a study unit (s) was determined as a percentage of the total hospital LOS. The Elixhauser Comorbidity Index was used for risk adjustment.26,27 Both observed and expected mortality were then weighted in proportion to the total LOS on a study unit, and the “weighted” risk-adjusted observed-to-expected (O:E)
JGIM
Braddock et al.: The TRANSFORM Patient Safety Project
Table 3. Safety Issues Discovered through Program Implementation and Actions Taken Teamwork Issues Suboptimal communication of changes in patient status
Inadequate bedside identification of caregiver roles Staff nurses not reliably informing charge nurses of patient condition changes Delays in communicating patient condition changes due to mobile nature of staff nurses on unit Early Detection Issues Failure to compare current status to baseline status Delays in diagnosis and treatment by residents due lack of consultation with senior clinician Variability in knowledge of diagnosis and treatment of study complications
Role of “relief nurse” on unit was task-driven Care Escalation Issues Least experienced resident being called first for urgent patient situation Suboptimal knowledge of chain of command (COC) and assertion in escalating care
Actions • 5-min videos demonstrating optimal Situation-Background-Assessment-Recommendation (SBAR) communication by unit clinical leaders was shown to staff during daily huddles • Charge nurses and patient safety champions were trained to “ask for an SBAR” from nurses when hearing of a condition change • A nursing competency on SBAR created requiring 1:1 demonstration and rated by CNSs • SBAR competency was observed during simulation training; if suboptimal, required to repeat performance • Paging guideline was implemented on study units to enhance use of SBAR when communicating patient problems • Extender placed on name badges identifying role (e.g., nurse, resident, attending physician) • Role identification was observed during simulation training and 1:1 feedback given during debriefing • Charge nurses began rounding on patients minimally once a shift to prompt staff nurses to relay clinical changes • Competency of having nurses communicate changes to charge nurses was observed during simulation training and feedback given • Nurses taught to relay phone number in patient’s room when paging resident to ensure a more timely response • Nursing units purchased individual phones for nurses to avoid delays in communicating problems to physicians • Communication behavior observed during simulation training and 1:1 feedback given during debriefing Actions • Nurses taught how to reconfigure electronic medical record to view baseline vital signs and laboratory values for comparison • Education on need for early consultation was reinforced during simulation training, patient safety conferences and monthly meetings • Consultation competency for nurses and residents observed during simulation training and 1:1 feedback given • Sepsis order set was created in electronic medical record to support optimal ordering of diagnostic tests and treatment • Clinical guidelines for study complications were routinely discussed in depth at quarterly patient safety conferences • Knowledge competencies and ordering behaviors were observed during simulation training and feedback given • Role description of relief nurse was revised to emphasize coaching role to help nurses critically think through condition changes and take action Actions • Unit guideline defined new process that intern and senior resident need to be concurrently called for urgent/emergent situations • Guideline disseminated among nurses and supported by UBMDs in monthly unit resident orientation • CNSs reinforced COC to staff nurses during huddles • Patient safety conferences reviewed COC and method for asserting care escalation in face of perceived authority gradients • Patient safety champions empowered staff nurses to elicit COC • COC knowledge and assertion competencies observed during simulation training and feedback given
methodology.28 Participant responses of “agree” and “strongly agree” on the survey’s five-point Likert scale constituted the mean percent positive scores (0–100 % measurement range). All survey results are reported as percent positive scores.
mortality ratio was determined, based on the premise that clinical deterioration of patients is proportional across inpatient units where they stayed during the LOS. Clinical outcomes were measured at baseline (July 2009–June 2010) and for intervention and sustainability periods. Safety culture on study units was assessed at baseline and after 1 year, using the Agency for Healthcare Research and Quality’s (AHRQ) Hospital Survey on Patient Safety Culture (HSOPS).28 HSOPS data for nurses on non-study, medicalsurgical inpatient units were available for baseline and postintervention periods from an organizational-wide survey. Survey results were analyzed following AHRQ scoring
Statistical Analysis Multivariate regression was used to assess the effects of the program on complication rates and mortality over the three time periods, adjusting for patient characteristics. Control variables included patient age, race, payer status, inpatient
Table 4. Rates of Hospital-Acquired Severe Sepsis/Septic Shock and Acute Respiratory Failure on Study Units
Rate of Severe Sepsis/Septic Shock (complications/1000 unit discharges) Cases of Severe Sepsis/Septic Shock Rate of Acute Respiratory Failure (complications/1,000 unit discharges) Cases of Acute Respiratory Failure Unit Discharges
Baseline Period (July 2009–June 2010)
Intervention Period (July 2010–June 2011)
Sustain Period (July–Dec 2011)
Odds Ratio (95 % CI)
p value
1.78
1.21
0.64
0.53 (0.29–0.96)
0.04
16 2.44
11 2.10
3 0.43
0.58 (0.35–0.96)
0.03
22 9,000
19 9,058
2 4,685
JGIM
Braddock et al.: The TRANSFORM Patient Safety Project
admission status, and Elixhauser comorbidities.26,27 Model parameters were estimated using unconditional least squares and goodness-of-fit evaluated by the likelihood ratio test. G charts were used to calculate the number of days between complications. Linear regression was used to evaluate the effect of the program on the rate of unplanned transfer to HLOC. Pearson’s chi-squared tests with Yates’ continuity correction and Fisher’s Exact test were used to analyze the safety culture data. All statistical analyses were performed in SAS v9.3 (SAS Institute, Inc., Cary NC, USA). A p value of < 0.05 was considered to denote statistical significance for all outcomes.
Microsystem leaders led 90 % of monthly patient safety team meetings during intervention and sustainability periods, and each study unit maintained one unit patient safety champion per shift. Patient safety conferences were held each quarter and 90 % of RRT and cardiopulmonary arrest calls were debriefed. Each month during the study, one nurse or resident received teamwork recognition. Of the numerous safety issues discovered by microsystem leaders during the study, the most prevalent issues focused on interdisciplinary teamwork, early detection and care escalation (Table 3).
Outcomes RESULTS
Program Interventions After 1 year of intervention, 90 % of nurses (N=247) and 92 % of residents (N=56) on study units had participated in at least one simulation training exercise, with 43 % of nurses and 20 % of residents participating in two or more. Four simulation exercises were conducted each month on each unit during the intervention period and one exercise was conducted each month on each unit during the sustainability period. At the end of the sustainability period, 98 % of nurses (N=269) and 100 % of residents (n=61) had participated in simulation training.
A total of 13,743 patients were discharged from the study units during the study. The rate of hospital-acquired severe sepsis/ septic shock on study units decreased from 1.78 to 0.64 per 1,000 unit discharges across intervention and sustainability periods (odds ratio, 0.53; 95 % CI, 0.29–0.96; p=0.04) (Table 4). The mean number of days between cases of severe sepsis/septic shock statistically increased from 18 at baseline to 34 during the intervention period (p=0.03), with rates decreasing to 25 during the sustainability period (Fig. 1a). No significant change in the mean number of days between cases of acute respiratory failure occurred across study periods (p=0.25) (Fig. 1b). The rate of hospital-acquired acute respiratory failure decreased from 2.44 to 0.43 across the study periods
Figure 1. G-Charts showing days between cases on study units of: a severe sepsis/septic shock, and b acute respiratory failure.
JGIM
Braddock et al.: The TRANSFORM Patient Safety Project
(odds ratio, 0.58; 95 % CI, 0.35–0.96; p=0.03) (Table 4). While rates on study units decreased over time, the rates of both complications occurring on non-study inpatient units statistically significantly increased during intervention and sustainability periods (Fig. 2a and b). The rate of unplanned transfers to HLOC increased from 715 to 764 per 1,000 unit transfers across the study periods (p=0.08) (Table 5a). The medical conditions primarily contributing to the increase in unplanned transfers across the study periods were conditions not related to study complications (Table 5b). The weighted, risk-adjusted observed-to-expected (O:E) mortality ratio on study units was 0.50 at baseline, which decreased to 0.44 during the intervention period, and decreased to 0.40 during the sustainability period (odds ratio, 0.95; 95 % CI, 0.94–0.97; p
